ARTICLE

Received 30 Dec 2016 | Accepted 20 Jan 2017 | Published 8 Mar 2017

MAX inactivation is an early event in GISTdevelopment that regulates p16 and cellproliferationInga-Marie Schaefer1, Yuexiang Wang1,w, Cher-wei Liang1,w, Nacef Bahri1, Anna Quattrone1,2, Leona Doyle1,

Adrian Marino-Enrıquez1, Alexandra Lauria1, Meijun Zhu1, Maria Debiec-Rychter2, Susanne Grunewald3,

Jaclyn F. Hechtman4, Armelle Dufresne1, Cristina R. Antonescu4, Carol Beadling5, Ewa T. Sicinska6,

Matt van de Rijn7, George D. Demetri8, Marc Ladanyi4, Christopher L. Corless5, Michael C. Heinrich9,

Chandrajit P. Raut10, Sebastian Bauer3 & Jonathan A. Fletcher1

KIT, PDGFRA, NF1 and SDH mutations are alternate initiating events, fostering hyperplasia in

gastrointestinal stromal tumours (GISTs), and additional genetic alterations are required for

progression to malignancy. The most frequent secondary alteration, demonstrated in B70%

of GISTs, is chromosome 14q deletion. Here we report hemizygous or homozygous

inactivating mutations of the chromosome 14q MAX gene in 16 of 76 GISTs (21%). We find

MAX mutations in 17% and 50% of sporadic and NF1-syndromic GISTs, respectively, and we

find loss of MAX protein expression in 48% and 90% of sporadic and NF1-syndromic GISTs,

respectively, and in three of eight micro-GISTs, which are early GISTs. MAX genomic

inactivation is associated with p16 silencing in the absence of p16 coding sequence deletion

and MAX induction restores p16 expression and inhibits GIST proliferation. Hence, MAX

inactivation is a common event in GIST progression, fostering cell cycle activity in early GISTs.

DOI: 10.1038/ncomms14674 OPEN

1 Department of Pathology, Brigham and Women’s Hospital, Harvard Medical School, 20 Shattuck Street, Thorn 528, Boston, Massachusetts 02115, USA.2 Department of Human Genetics, KU Leuven and University Hospitals Leuven, Herestraat 49, Box 602, B-3000 Leuven, Belgium. 3 Sarcoma Center, WesternGerman Cancer Center, University of Duisburg-Essen Medical School, Hufelandstrasse 55, 45122 Essen, Germany. 4 Department of Pathology, MemorialSloan-Kettering Cancer Center, 1275 York Avenue, New York, New York 10065, USA. 5 Department of Pathology, Knight Cancer Institute, Oregon Health andScience University, 3181 Southwest Sam Jackson Park Road, Portland, Oregon 97239-3098, USA. 6 Department of Oncologic Pathology, Dana-Farber CancerInstitute, Harvard Medical School, 450 Brookline Avenue, Boston, Massachusetts 02215, USA. 7 Department of Pathology, Stanford University MedicalCenter, 300 Pasteur Drive, Stanford, California 94305, USA. 8 Ludwig Center at Harvard, Harvard Medical School and Department of Medical Oncology,Dana-Farber Cancer Institute, 450 Brookline Avenue, Boston, Massachusetts 02215, USA. 9 Portland VA Health Care System, Knight Cancer Institute, OregonHealth and Science University, 3181 Soutwest Sam Jackson Park Road, Portland, Oregon 97239-3098, USA. 10 Department of Surgery, Brigham and Women’sHospital, Harvard Medical School, Boston, 75 Francis Street, Boston, Massachusetts 02115, USA. w Present addresses: Changzheng Hospital Joint Center forTranslational Medicine, Institutes for Translational Medicine (CAS-SMMU); Key Laboratory of Stem Cell Biology, Institute of Health Sciences, SIBS, ChineseAcademy of Sciences, Shanghai JiaoTong University School of Medicine; Collaborative Innovation Center of Systems Biomedicine, 320 Yueyang Road,Shanghai 200025, China (Y.W.); Department and Graduate Institute of Pathology, National Taiwan University Hospital and National Taiwan UniversityCollege of Medicine, 7 Zhong-Shan South Road, Taipei, Taiwan 10002 (C.-w.L.). Correspondence and requests for materials should be addressed to J.A.F.(email: jfletcher@partners.org).

NATURE COMMUNICATIONS | 8:14674 | DOI: 10.1038/ncomms14674 | www.nature.com/naturecommunications 1

Activating mutations of the receptor tyrosine kinases KIT1

or PDGFRA2 are initiating or early events in mostgastrointestinal stromal tumours (GISTs) and indeed are

present in micro-GISTs, which are asymptomatic subcentimetreGIST lesions found in one-third of the general population3.Genetic progression from micro-GIST to malignant GIST resultsfrom stepwise accumulation of deletions in chromosome arms14q, 22q, 1p and 15q, together with cell cycle dysregulatingevents and dystrophin inactivation4–8. These highly recurrentchromosomal deletions implicate losses of yet-unidentifiedtumour suppressor mechanisms in GIST progression. Of these,the 14q deletions are most frequent, observed in 60–70% of GISTs(including neurofibromatosis type 1 (NF-1)-associated GIST) asan early event in genetic progression4–6,9,10.

Here we show that 14q deletions target the MAX transcrip-tional regulator gene in early GISTs of various molecular origins(KIT-mutant, PDGFRA-mutant or NF1-mutant). These MAXgenomic-inactivating mutations are driver events, enabling GISTprogression by loss of MAX expression, and consequent p16silencing and cell cycle dysregulation.

ResultsGenomic studies. Targeted sequencing of 812 cancer-associatedgenes demonstrated somatic homozygous inactivating MAXmutations in three of ten GISTs (Supplementary Data 1 andSupplementary Fig. 1a). The ten GISTs in this discovery set hadKIT mutations (seven cases), PDGFRA mutations (two cases) andNF1 mutation (one case) (Supplementary Data 1). Apartfrom KIT and PDGFRA, MAX was the only other gene withdemonstrable recurrent mutations in this discovery set. MAXevaluations by Ion AmpliSeq sequencing, Sanger sequencing(Supplementary Fig. 1b), quantitative PCR (SupplementaryFig. 1c) and single-nucleotide polymorphism (SNP) arrays(Supplementary Fig. 1d) were performed in the same 10 GISTsand in 66 additional GISTs (Supplementary Data 2). This total setof 76 GISTs was shown to have mutually exclusive mutationsinvolving the KIT, PDGFRA, NF1 and SDH genes in 52 (68%), 8(11%), 11 (14%) and 2 (3%) cases, respectively (SupplementaryData 2). These assays demonstrated somatic hemizygous orhomozygous MAX-inactivating mutations in 16 of the 76 GISTs(21%), including 8 mononucleotide mutations and 8 larger-scale intragenic deletions (Fig. 1 and Supplementary Data 2).Non-neoplastic companion DNAs were MAX wild type for sevenof eight GISTs with mononucleotide MAX mutations, showingthat these were somatic mutations, and the MAX mutation allelicfrequency in the remaining case (case 19) was 0.7, consistent witha hemizgyous or homozygous somatic mutation. The MAXmononucleotide mutations were nonsense (N¼ 3), loss of startcodon (N¼ 1), frame-shift (N¼ 1), splice site (N¼ 2) and50-untranslated region (N¼ 1). The two splice-site mutations

(cases 7 and 59) destroyed invariant splicing motifs, creatinginactive MAX transcripts with loss of exon 3 (case 7) or retentionof intron 4 (case 59), as confirmed by reverse transcriptase–PCR,and—for case 7—also confirmed by genome RNA sequencing(Supplementary Fig. 2). The 50-untranslated region mutation(case 19) was predicted to be functionally relevant by thePROMO.3 tool, with predicted disruption of transcription factorbinding sites in the MAX promoter. Multiple anatomicallydistinct specimens were studied in five patients (pts) withMAX-mutant GISTs and all had identical mutations, as shown bycomparisons of primary GIST and subsequent metastases in twopts, and by comparisons of multiple metastases (two–tenmetastases analysed per pt) in three pts (Supplementary Data 2and Supplementary Fig. 3).

Among the overall study group of 76 GISTs, the GIST primarysite was known for 71 pts, whereas primary site could not bedetermined in the remaining 5 pts who presented withdisseminated intra-abdominal disease. MAX genomic mutationswere more common in non-gastric than gastric GISTs (P¼ 0.001for the 71 pts with known GIST primary sites, two-tailed Fisher’sexact test) and this association remained significant when GISTswith NF1 mutations—which are generally of small bowel origin—were removed from consideration (P¼ 0.004 for 61 pts with NF1-wild type GISTs of known primary sites).

Protein studies. MAX was assessed in each of the 76 GISTs byimmunoblotting (N¼ 75) and/or immunohistochemistry (IHC)(N¼ 22). MAX inactivation was demonstrated in 38 of 75 GISTs(51%) by immunoblotting (Fig. 2) and was associated with MAXgenomic mutation (Po0.0001, two-tailed Fisher’s exact test).Likewise, MAX inactivation, as demonstrated by IHC in 14 of 22GISTs (63%) (Fig. 3), was associated with MAX genomic inacti-vation (P¼ 0.0062). Loss of MAX expression, when detected inany GIST metastasis, was detected in all others from the same pt(Supplementary Data 2 and Supplementary Fig. 4). Forty per centof the GISTs with loss of MAX expression were classified,according to well-established clinicopathological criteria11,as ‘low-risk’ and ‘intermediate-risk’ cases (Fig. 4), which arestages of GIST development that precede transition to clinicallyaggressive ‘high-risk’ GIST. These findings show that MAXinactivation can be an early event in GIST biological and clinicalprogression. Further, MAX inactivation was detected in threeof eight micro-GISTs, each of which had 14q deletion(Supplementary Fig. 5), confirming MAX dysregulation as anearly event in GIST progression (Supplementary Table 1).Hemizygous or homozygous inactivating MAX mutations weredemonstrated in sporadic GISTs (11 of 64¼ 17%) and insyndromic GISTs in individuals with NF-1 (5 of 10¼ 50%)(Fig. 4 and Supplementary Data 2). Likewise, loss of MAX proteinexpression was demonstrated in both sporadic (31 of 64¼ 48%)

MAX

SNVs(n=8)

Homdeletions(n=8)

1 1,546 bp

ATG TAAE15’-UTR 3’-UTRE2 E3 E4 E5

c.-94G>C IVS2-2A>G IVS4+1G>Ac.97C>T c.145C>A c.228delT c.289C>Tc.3G>A

p.M1I p.R33* p.S49* p.Y70fs p.Q97*

Figure 1 | Genomic MAX mutations in 76 GISTs. Inactivating MAX mutations were intragenic homozygous deletions (blue lines indicate deleted exons)

and hemizygous mononucleotide alterations. Mutations are described according to international guidelines for sequence variant nomenclature by the

Human Genome Variation Society (http://varnomen.hgvs.org). Annotations in blue are the nucleotide coding sequence mutations (indicated by ‘c.’),

whereas annotations in green are the resultant protein sequence mutations (indicated by ‘p.’). All mutations affect both alternatively spliced forms of MAX,

which encode 151 and 160 amino acid MAX isoforms.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14674

2 NATURE COMMUNICATIONS | 8:14674 | DOI: 10.1038/ncomms14674 | www.nature.com/naturecommunications

and NF-1-associated GISTs (9 of 10¼ 90%). GISTs occur in up to25% of NF-1 pts, predominantly in the small intestine and oftenas multicentric tumours12–14. Our studies credential MAX

inactivation, as an early step in GIST progression, associatedwith KIT and PDGFRA gain-of-function mutations and NF1 loss-of-function mutations.

Functional studies. The GIST48 cell line has MAX inactivationdue to homozygous deletion of MAX exons 1 and 2, and has lossof p16 (p16INK4A) expression. This cell line shows a localizedCDKN2A deletion, which affects the p14ARF coding sequence(Supplementary Fig. 6) but lacks genomic alterations of the p16coding sequence in CDKN2A and lacks CDKN2A methylation.CDKN2A ranked in the top 0.1% of genes differentially expressedafter MAX restoration in GIST48 and was the highest rankingcancer-associated gene (Supplementary Data 3). In keeping withthis evidence, p16 protein expression was strong and diffuse inGISTs lacking MAX or p16INK4A coding sequence mutations,but was undetectable in MAX-mutant GISTs, even in the absenceof p16 coding sequence mutation (Supplementary Fig. 7). Indu-cible restoration of MAX expression in GIST48 upregulatedCDKN2A transcript expression, restored p16 protein expressionand inhibited RB1 phosphorylation (pRB1Ser795) (Fig. 5a) and cellproliferation (Fig. 5b,c). MAX restoration in GIST48 was notassociated with significant enrichment of MYC-related expressionsignatures or with altered sensitivity to MYC:MAX inhibitordrugs (Supplementary Fig. 8), suggesting that MAX tumoursuppressor roles in GIST are not necessarily MYC dependent.

DiscussionAll MAX-inactivating point mutations in this series werehemizygous, with allelic frequencies typically B0.67 in GISTsknown to have loss of the other MAX allele and containingB20% non-neoplastic cells. Similarly, all MAX intragenicdeletions were homozygous. These findings indicate that GISTprogression is best served by complete loss of MAX function.Although the Ion AmpliSeq and HaloPlex assays detect Z5%mutant alleles15,16, we found no evidence for multiple

MAX

LRisk:

a

b

Risk:

L I I H M M MW (kDa)

L L L L H H H

Sporadic GIST

NF-1 associated GIST

GAPDH

GAPDH

1 0 1.02 0.28

GIS

T430

GIS

T48

GIS

T430

GIS

T48

0.57 0.32 0.33 0.44 0.12

1 0 0.19 0.08 0.08 0.2 0.36 0.03 0.05

MAX

24

38

24

38

Figure 2 | Loss of MAX expression in both sporadic and NF-1-associated

syndromic GISTs. MAX protein inactivation is demonstrated by

immunoblotting GIST snap-frozen biopsies from sporadic (a) and NF-1-

associated (b) cases. MAX wild-type GIST430 cell line and MAX-mutant

GIST48 cell line are positive and negative controls, respectively. MAX

inactivation was defined by expression level o0.4, normalized to GIST430.

‘L’ denotes low risk, ‘I’ intermediate risk, ‘H’ high risk and ‘M’ metastatic.

The cases, from left to right in a are 21, 22, 32, 3, 39, 58 and 53, and in b are

73, 68, 71, 74, 76, 75 and 7. Residual MAX expression in MAX-mutant

cases (53, 73, 68, 76 and 7) results from non-neoplastic cells (fibroblasts,

endothelial cells and inflammatory cells) and from admixed precursor GIST

cells that had not yet acquired the MAX mutations.

a b c d

e f g h

Figure 3 | Loss of MAX protein expression can be present in cell subpopulations in early GISTs but is present in all cells from affected metastatic

GIST. Haematoxylin and eosin stains (a,c,e,g) and MAX IHC (b,d,f,h): case 67 with wild-type MAX (a,b) has retained MAX expression; case 29 (c,d), low-risk

GIST with MAX mutation, has diffuse loss of MAX expression; case 3 (e,f), intermediate-risk GIST, has mosaic loss of MAX expression with admixed MAX-

positive and MAX-negative cells; case 51 (g,h), metastatic GIST with MAX mutation, has diffuse loss of MAX expression. Positive internal controls for MAX

expression in all cases are scattered inflammatory cells and fibrovascular cells, and—in case 51—a lymphoid aggregate at the upper left. Scale bar, 50 mm.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14674 ARTICLE

NATURE COMMUNICATIONS | 8:14674 | DOI: 10.1038/ncomms14674 | www.nature.com/naturecommunications 3

MAX-mutant subclones within primary GISTs or betweendifferent metastases in a given pt. These observations indicatethat a single KIT/PDGFRA-mutant/MAX-mutant or NF1-mutant/MAX-mutant subclone fosters biological and clinical progressionin MAX-mutant GISTs. MAX is a helix-loop-helix leucine zippertranscription factor, which regulates cell proliferation,differentiation and apoptosis through heterodimerization withMYC-family proteins17–19, but MAX homodimers also regulatetranscription in a MYC-independent manner20,21. MAX has

tumour suppressor roles in a small subset of hereditarypheochromocytomas and in small cell lung cancer21,22, and arecent report demonstrated MAX mutation in one KIT/PDGFRAwild-type, NF1-mutant, GIST23. Our study demonstrates thatMAX mutations are common alterations in GIST, occurring atearly stages of GIST development. We detected hemizygous andhomozygous MAX-inactivating mutations in 21% of the 76 GISTsin this study, thereby demonstrating that GIST is the neoplasiawith the highest known frequency of MAX tumour suppressor

N

SDH deficientS

Risk:

GIST genotype:

MAX genomic:

MAX protein:

Low risk

Intermediate risk

Metastatic

High risk

Risk:

GIST genotype:

MAX genomic:

MAX protein:

1 10 20 30

40 50 60 70

L L L L L

L

L L L L L L L L L L L L L L L L L L L L L L L L I I I I I I I

K K K K K K K K K K K K K K K P P P P N N N N N N N S S K K K K K P P

K K K K K K K K K K PK P N N N N K K K K K K K K K K K K K K K K K K K K K

I

KIT mutant

PDGFRA mutant

NF1 mutant

MAX mutation Loss of MAX expression

H

H

H H H H H H H H H H H H H H H H H M M M M M M M M M M M M M M M M M M M M M M

M

K

P

Figure 4 | Summary of MAX genomic and protein aberrations in 76 GISTs. Results are shown for both early and late stages of GIST progression (risk

classifications), and for GISTs initiated by KIT, PDGFRA, NF1 and SDH mutations.

0 24 48 72 24

GIST48

GIST48MAXGIST48

GIST48MAX

0

2

4

6

8

10

10

15

0

5

Cel

l num

bers

(x1

05 )

Brd

U in

corp

orat

ion

(x10

6 )

Time (h) Time (h)

*

*

**

* P = 0.0130* P = 0.0488

** P = 0.0029

b c

a

GIST48Lo

w-risk

GIS

T

GIST48

GIST48

MAX

GIST48MAXDOXY: MW (kDa)

24

17

102

102

38

+ +–

107.7 62.5

8.8 7.2

1.1 1.1

GAPDH

RB1

Phospho-RB1 S795

p16

MAX0 2.8

0 0.6

GAPDH

RB1

Phospho-RB1 S795

p16

MAX

0% 50% 100%

Figure 5 | MAX restoration also restores p16 expression and function in GIST. MAX restoration in MAX-mutant (homozygous deletion) GIST48 cell line

restores p16 expression and inhibits CDK4/6-dependent phospho-RB1Ser795; GAPDH serves as loading control and the bar graph normalizations of GIST48

versus GIST48MAX are with the higher value set to 100% (a). Restoration of MAX expression inhibits GIST48 growth, as assessed by cell counts at 24 h

(P¼0.0488) and 72 h (P¼0.0029) (b), and inhibits GIST proliferation index, as assessed by BrdU incorporation at 24 h (P¼0.0130) (Student’s t-test)

(c). Tests were performed in triplicate. The error bars show s.d. (b) and s.e.m. (c).

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14674

4 NATURE COMMUNICATIONS | 8:14674 | DOI: 10.1038/ncomms14674 | www.nature.com/naturecommunications

mutations. Further, the true frequency of MAX mutations inGIST is likely to be 421%, given that homozygous deletions werea common mechanism of MAX inactivation. Homozygousdeletions, particularly if small, will be difficult to detect in someearly GISTs, where they are present in a subclone of the overallneoplastic proliferation.

We show MAX protein inactivation in B50% of GISTs,including micro and low-risk GISTs, which is additional evidencefor MAX dysregulation as an early event in GIST development.MAX restoration in GIST48 cells inhibited GIST cell growth andupregulated CDKN2A expression, accompanied by p16 upregula-tion and inhibition of pRB1Ser795. These findings suggest thatMAX inactivation causes cell cycle dysregulation at an early pointin GIST progression, probably enabling progression to GISTstages with greater proliferative potential. Altogether, our studiesdemonstrate frequent disruption of MAX tumour suppressiveroles during early progression of KIT-mutant, PDGFRA-mutantand NF1-mutant GISTs.

MethodsTumour and tissue samples. Discarded, de-identified tumour specimens wereobtained at Brigham and Women’s Hospital and Memorial Sloan-Kettering CancerCenter, under protocols approved by the Dana-Farber/Brigham and MemorialSloan-Kettering Cancer Center Institutional Review Boards. Informed writtenconsent was obtained from all human participants.

Cell lines. GIST cell lines were established in the Fletcher laboratory and werevalidated against the initial biopsy material by molecular cytogenetics andsequencing verification of known unique gene mutations. Daudi cells wereobtained from ATCC (Manassus, VA, USA). All cultures were shown to bemycoplasma free.

Targeted sequencing. Targeted sequencing was performed using the HaloPlexTarget Enrichment System for Illumina sequencing (Agilent Technologies, SantaClara, CA, USA) for a custom-designed set of 812 cancer-related genes. Datawere analysed by SureCall software (version 2.0.7.0, Agilent Technologies) andIntegrative Genomics Viewer (IGV) (version 2.3.25, Broad Institute). Targetedhighly multiplexed PCR with semiconductor-based sequencing using the IonAmpliSeq assay was performed as described previously15, analysing the MAX gene(NM_002382) coding sequence, an additional two nucleotides adjacent to eachexon, and 1 kb of upstream sequence. Amplicon size ranged from 125 to 275 bp(including primers) with an average of 243 bp. Inserts ranged in size from 77 to230 bp (excluding primers), with an average of 194 bp. Ion AmpliSeq detection forhomozygous deletions was performed after normalization to non-neoplastic DNAsequences and establishing cutoffs based on estimated presence of 30% non-neoplastic cells in low-/intermediate-risk GISTs and 20% non-neoplastic cells inhigh risk/metastatic GISTs. Deletion of at least nine consecutive amplicons and/ora ratio of o0.4 for markers located at either the 30- or 50-end of the gene in relationto all markers in a given case were defined as criteria for homozygous deletion.

PCR analysis. Genomic PCR and Sanger sequencing of MAX exons 1–5(NM_002382) was performed using primers shown in Supplementary Table 2.Reverse transcription from tumour RNA was performed using the iScript cDNASynthesis kit (Bio-Rad Laboratories, CA) and followed by PCR using PlatinumPCR Super Mix (Life Technologies, Grand Island, NY, USA) with analysis bySanger sequencing.

Quantitative PCR. Quantitative PCR of MAX exons 1, 3 and 4, and the flankinggenes FNTB and FUT8 was performed using 25ml volume per reaction, containing5 ng of genomic DNA, 100 nM forward primer, 100 nM reverse primer and 1� iQSYBR Green Supermix (Bio-Rad). The MyiQ single colour real-time detectionsystem (Bio-Rad) was used for thermal cycling. Samples were run in triplicate withnon-template control, human non-neoplastic cell DNA and GIST48 DNA with aknown homozygous MAX deletion. Mixtures of non-neoplastic and MAX-mutantGIST48 DNA (30:70 and 20:80) were controls modelling detection of MAXhomozygous deletions in human GIST biopsies (in which 10–30% of cells arenonneoplastic). Amplification accuracy was verified by melting curve analysis. Theminimum threshold cycle (Ct value) generated by the MyiQ software (Bio-Rad) foreach sample was used to calculate MAX copy number using Ct values for normaltissue and adjacent genes as reference.

SNP arrays. High-molecular-weight gDNA was isolated using QIAamp DNAMini Kit (Qiagen, Valencia, CA, USA) and analysed by Affymetrix Cytoscan HD

2,600 K SNP array (Affymetrix, Santa Clara, CA, USA) with AffymetrixChromosome Analysis Suite 2.0.

Array comparative genomic hybridization. Micro-GIST gDNAs were obtainedfrom microdissected tissues and extracted using QIAamp DNA FFPE Tissue kits(Qiagen) following the manufacturer’s protocol. Amplification was performedusing a GenomePlex Tissue Whole Genome Amplification WGA5 kit (Sigma, SaintLouis, MO, USA). The post-WGA products were purified using a QIAquickPCR Purification Kit (Qiagen) and quantified using a NanoDrop ND-1000Spectrophotometer. Comparative genomic hybridization was performed using acustomized human array comparative genomic hybridization platform with2� 415 K coverage (Agilent Technologies). Four independent experiments wereconcurrently performed per template amplification and then mixed, to minimizeamplification bias and allele dropout. Data were analysed using Agilent Technol-ogies 10.5.1.1 Software.

Protein blotting. Whole-cell lysates were prepared as described previously24

and protein concentrations were determined using the Bio-Rad protein assay(Bio-Rad). Electrophoresis, immunoblotting and chemiluminescence detectionwere as described previously8. Primary antibodies were directed against MAX(Santa Cruz Biotechnology, Dallas, TX, USA, C-17, 1:200 dilution), p16INK4A/CDKN2A (R&D Systems, Minneapolis, MN, USA, AF5779, 1:200 dilution),pRB1Ser795 (Cell Signaling Technology, Danvers, MA, USA, 9301, 1:1,000 dilution)and GAPDH (Sigma, GAPDH-71.1, 1:5,000 dilution). Full-length blottings can beviewed in Supplementary Fig. 9.

Immunohistochemistry. Immunohistochemical staining for MAX was performedwith the MAX C-17 antibody (Santa Cruz) at dilution of 1:1,500 on 4 mm thinsections prepared from formalin-fixed, paraffin-embedded tissue blocks afterantigen retrieval using a citrate buffer pressure cooker protocol. Staining for p16(p16INK4A) was performed using a mouse monoclonal antibody (dilution 1:2;clone E6H4, Ventana Medical Systems, Tucson, AZ, USA).

MAX restoration. Lentivirus preparations were produced by cotransfecting MAXconstruct (Broad Institute, clone ID BRDN0000560330, NM_002382.3) introducedinto a destination vector pLXI_TRC401 (Alias pCW57.1 Dest, TRE-Gateway-NoTag) by LR Clonase (Thermo Fisher Scientific, Waltham, MA, USA) reaction andhelper virus packaging plasmid pCMVDR8.91 and pMD.G into 293T cells, asdescribed previously25. Lentivirus was harvested at 24, 36, 48 and 60 h posttransfection and virus titres were quantified and stored at � 80 �C. GIST48transductions were carried out overnight with polybrene 8 mg ml� 1 (Sigma) andtransduced cells were selected for 9 weeks with puromycin (0.125 mg ml� 1), whichwas discontinued 7–10 days before analyses. MAX expression was induced in stablytransduced cells with doxycycline (2.5 mg ml� 1) every 36 h.

Gene expression profiling. RNA sequencing was performed 24 h after MAXrestoration in GIST48 cells using an Illumina HiSeqTM 2000 platform (BeijingGenomics Institute, Hong Kong). Parental GIST48 cells treated with doxycyclinefor 24 h served as control. Data analyses were with BGI-Tech Pipeline Version 3.1and differentially expressed genes were identified using criteria false discovery rater0.001 and abs(log2(MAX-/MAXþ ))Z1.

BrdU uptake and CellTiter-Glo analyses. Cells were plated in 96-well plates at20,000 cells per well in growth medium and incubated overnight. Cells were treatedwith 2.5 mg ml� 1 doxycycline for 24 h (MAX-restoration) and parental GIST48was the untreated comparator. For 5-bromodeoxyuridine (BrdU) incorporationproliferation analyses, BrdU was added to the cells for 24 h. BrdU incorporation,fixation and detection were performed using a BrdU Cell Proliferation ELISA asper the manufacturer’s protocol (Roche Diagnostics, Indianapolis, IN, USA).BrdU incorporation was presented as % of untreated control. For CellTiter-Glo(Promega, Madison, WI, USA) viability analyses, MYC inhibitors versus dimethylsulfoxide-only control were added for 3 days and ATP incorporation was thenmeasured using a luminometer. All cell response assays were performed intriplicate wells, with the entire study replicated at least once.

Cell counts. Cells were trypsinized, resuspended in media and counted byhaemocytometer.

Statistical analyses. Statistical analyses were performed using GraphPad PrismSoftware and Student’s t-test and two-tailed Fisher’s exact tests to compare twodata sets.

Data availability. Haloplex targeted DNA sequencing data of the discoverycohort, Ion Ampliseq MAX sequencing data of all cases and gene expression datahave been deposited in the Sequence Read Archive (accession number SRP096291).Cytoscan HD SNP array data have been deposited in the Gene Expression

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14674 ARTICLE

NATURE COMMUNICATIONS | 8:14674 | DOI: 10.1038/ncomms14674 | www.nature.com/naturecommunications 5

Omnibus database (accession number GSE93077). All remaining data are availablewithin the article and Supplementary Information files or available from theauthors upon request.

References1. Hirota, S. et al. Gain-of-function mutations of c-kit in human gastrointestinal

stromal tumors. Science 279, 577–580 (1998).2. Heinrich, M. C. et al. PDGFRA activating mutations in gastrointestinal stromal

tumors. Science 299, 708–710 (2003).3. Agaimy, A. et al. Minute gastric sclerosing stromal tumors (GIST tumorlets)

are common in adults and frequently show c-KIT mutations. Am. J. Surg.Pathol. 31, 113–120 (2007).

4. El-Rifai, W., Sarlomo-Rikala, M., Andersson, L. C., Knuutila, S. & Miettinen, M.DNA sequence copy number changes in gastrointestinal stromal tumors: tumorprogression and prognostic significance. Cancer Res. 60, 3899–3903 (2000).

5. Wozniak, A. et al. Array CGH analysis in primary gastrointestinal stromaltumors: cytogenetic profile correlates with anatomic site and tumoraggressiveness, irrespective of mutational status. Genes Chromosomes Cancer46, 261–276 (2007).

6. Gunawan, B. et al. An oncogenetic tree model in gastrointestinal stromaltumours (GISTs) identifies different pathways of cytogenetic evolution withprognostic implications. J. Pathol. 211, 463–470 (2007).

7. Henze, J. et al. p53 modulation as a therapeutic strategy in gastrointestinalstromal tumors. PLoS ONE 7, e37776 (2012).

8. Wang, Y. et al. Dystrophin is a tumor suppressor in human cancers withmyogenic programs. Nat. Genet. 46, 601–606 (2014).

9. Schaefer, I. M. et al. Chromosomal aberrations in primary PDGFRA-mutatedgastrointestinal stromal tumors. Hum. Pathol. 45, 85–97 (2014).

10. Yamamoto, H. et al. Neurofibromatosis type 1-related gastrointestinal stromaltumors: a special reference to loss of heterozygosity at 14q and 22q. J. CancerRes. Clin. Oncol. 135, 791–798 (2009).

11. Miettinen, M. & Lasota, J. Gastrointestinal stromal tumors: pathology andprognosis at different sites. Semin. Diagn. Pathol. 23, 70–83 (2006).

12. Agaimy, A., Vassos, N. & Croner, R. S. Gastrointestinal manifestations ofneurofibromatosis type 1 (Recklinghausen’s disease): clinicopathological spectrumwith pathogenetic considerations. Int. J. Clin. Exp. Pathol. 5, 852–862 (2012).

13. Miettinen, M., Fetsch, J. F., Sobin, L. H. & Lasota, J. Gastrointestinal stromaltumors in patients with neurofibromatosis 1: a clinicopathologic and moleculargenetic study of 45 cases. Am. J. Surg. Pathol. 30, 90–96 (2006).

14. Mussi, C., Schildhaus, H. U., Gronchi, A., Wardelmann, E. & Hohenberger, P.Therapeutic consequences from molecular biology for gastrointestinal stromaltumor patients affected by neurofibromatosis type 1. Clin. Cancer Res. 14,4550–4555 (2008).

15. Beadling, C. et al. Combining highly multiplexed PCR with semiconductor-based sequencing for rapid cancer genotyping. J. Mol. Diagn. 15, 171–176(2013).

16. Stoddard, J. L., Niemela, J. E., Fleisher, T. A. & Rosenzweig, S. D. Targeted NGS:a cost-effective approach to molecular diagnosis of PIDs. Front Immunol. 5, 531(2014).

17. Blackwood, E. M. & Eisenman, R. N. Max: a helix-loop-helix zipper protein thatforms a sequence-specific DNA-binding complex with Myc. Science 251,1211–1217 (1991).

18. Prendergast, G. C., Lawe, D. & Ziff, E. B. Association of Myn, the murinehomolog of max, with c-Myc stimulates methylation-sensitive DNA bindingand ras cotransformation. Cell 65, 395–407 (1991).

19. Blackwood, E. M., Luscher, B. & Eisenman, R. N. Myc and Max associatein vivo. Genes Dev. 6, 71–80 (1992).

20. Nair, S. K. & Burley, S. K. X-ray structures of Myc-Max and Mad-Maxrecognizing DNA. Molecular bases of regulation by proto-oncogenictranscription factors. Cell 112, 193–205 (2003).

21. Comino-Mendez, I. et al. Exome sequencing identifies MAX mutations as acause of hereditary pheochromocytoma. Nat. Genet. 43, 663–667 (2011).

22. Romero, O. A. et al. MAX inactivation in small cell lung cancer disruptsMYC-SWI/SNF programs and is synthetic lethal with BRG1. Cancer Discov. 4,292–303 (2014).

23. Belinsky, M. G. et al. Somatic loss of function mutations in neurofibromin 1and MYC associated factor X genes identified by exome-wide sequencing in awild-type GIST case. BMC Cancer 15, 887 (2015).

24. Rubin, B. P. et al. KIT activation is a ubiquitous feature of gastrointestinalstromal tumors. Cancer Res. 61, 8118–8121 (2001).

25. Ou, W. B., Zhu, M. J., Demetri, G. D., Fletcher, C. D. & Fletcher, J. A. Proteinkinase C-theta regulates KIT expression and proliferation in gastrointestinalstromal tumors. Oncogene 27, 5624–5634 (2008).

AcknowledgementsWe thank Raymond DiDonato, PhD, from Agilent Technologies, for HaloPlex targetedcancer gene panel design and for other helpful advice with the HaloPlex analyses, andJason Hornick, MD, PhD for assistance with the p16 IHC studies. This work wassupported by grants from the US National Institutes of Health, including 1P50CA127003(J.A.F. and G.D.D.), 1P50CA168512 (J.A.F. and A.M.-E.), the GIST Cancer ResearchFund (J.A.F., M.C.H., C.L.C., C.R.A.), the Life Raft Group (J.A.F., M.v.d.R., M.C.H.,C.L.C, S.B.), the Deutsche Krebshilfe (I.-M.S.), the VA Merit Review Grants1I01BX000338-01 and 2I01BX000338-05 (M.C.H.) and the V Foundation TranslationalResearch Grant (M.C.H.).

Author contributionsJ.A.F. supervised the research. J.A.F., I.-M.S., Y.W. and C.-w.L. conceived and designedthe experiments. I.-M.S., Y.W., C.-w.L., N.B., A.Q., L.D., A.L., M.Z., S.G., J.F.H., C.B.,M.L., C.L.C., M.C.H. and S.B. performed the experiments. J.A.F., A.M.-E., I.-M.S. and C.-w.L. performed statistical analysis. I.-M.S., Y.W., C.-w.L., N.B., A.Q., A.M.-E., S.G., J.F.H.,A.D., C.B., M.v.d.R., M.L., M.C.H., S.B. and J.A.F. analysed the data. C.-w.L., C.R.A.,M.v.d.R., M.L., E.T.S., C.P.R., S.B. and J.A.F. contributed reagents, materialsand /or analysis tools. J.A.F. and I.-M.S. wrote the paper. M.D.-R., G.D.D. and C.L.C.provided scientific advice and helpful comments on the project. All authors read andapproved the final manuscript.

Additional informationSupplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Competing financial interests: C.L.C. has received consulting fees from ThermoFisher.The remaining authors declare no competing financial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

How to cite this article: Schaefer, I.-M. et al. MAX inactivation is an early eventin GIST development that regulates p16 and cell proliferation. Nat. Commun. 8, 14674doi: 10.1038/ncomms14674 (2017).

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.

This work is licensed under a Creative Commons Attribution 4.0International License. The images or other third party material in this

article are included in the article’s Creative Commons license, unless indicated otherwisein the credit line; if the material is not included under the Creative Commons license,users will need to obtain permission from the license holder to reproduce the material.To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

r The Author(s) 2017

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14674

6 NATURE COMMUNICATIONS | 8:14674 | DOI: 10.1038/ncomms14674 | www.nature.com/naturecommunications